© 1989 Nature Publishing Group NEWS AND VIEWS SENSORY-MOTOR CONTROL------------------------------------------------------ Listening to the voice within Alan D. Grinnell LIKE most types of motor behaviour, vocalization depends critically on sensory feedback. Yet almost nothing is known of how or where this sensory-motor integra- tion takes place in mammals. It is very satisfying, therefore, to see the report by Metzner on page 529 of this issue 1 , based on extensive single-unit recording from awake, spontaneously-vocalizing mam- mals, that identifies a site and provides a cellular mechanism for auditory feedback control of one of the most precisely regu- lated vocalization behaviours known: Doppler-shift compensation in echo- locating bats. Several species of insectivorous bats use echo-location sounds composed of a 10-100-ms-long constant-fre- quency component, terminating in a brief, downward frequency sweep. The frequency of the con- stant component (approximately 78 kHz in the rufous horseshoe bats, Rhinolophus rouxi, used in this study) is remarkably accurately regulated, to less than ± 50 Hz, for sounds emitted by a particular bat at rest. Echoes of sounds emit- ted in flight return Doppler-shifted by up to 4-5 kHz, depending on the relative velocity between the bat and target. Bats use the small frequency and intensity fluctua- tions in the echo, due for example to insect wing movements, to detect and identify targets'. Two remarkable and well- documented adaptations make this possible. First, the auditory ner- vous system is sharply tuned to a frequency very close to that of the constant frequency at rest, so that small changes in echo frequency cause large changes in response'· 4 • Second, to take advantage of this sharp tuning, the bat lowers the frequency of its next emis- sion just enough to compensate for the upward Doppler-shift in the most recent echo, so that the echo is always held close to the region of maximum auditory sensi- tivity5. This Doppler-shift compensation can be as accurate as± 50-100Hz, repre- senting a sensory-motor regulation to within about 0.1 per cent of the constant frequency, over a frequency range of 5 per cent or more. Few, if any, more accurate forms of sensory-feedback regulation exist. Previous research has yielded clues to where one might look for auditory feed- back control of vocalization. The con- stant-frequency component of the emitted sound is controlled by the degree of contraction of the cricothyroid muscle; this, in turn, is directly correlated with the 488 discharge rate of the superior laryngeal nerve arising in the nucleus ambiguus in the brainstem'. Ablation experiments indicate that the behaviour of Doppler-shift com- pensation can be controlled at the midbrain and tegmental levels'·', and there have been reports of attenuation of evoked or single- unit responses in the midbrain and cortical levels during vocalization in bats and primates 9 · 10 • Most interesting is Schuller's description 11 of four units in the inferior colliculus of R. ferrum-equinum that respond differently or respond only to a sound when it is coupled to an electrically elicited vocal- ization. None of these studies, however, reveals behaviour that helps explain their Rhino/ophus rouxi- flying blind. sensory-feedback control of vocalization. Metzner 1 has now surveyed single-unit behaviour in much of the midbrain and brainstem in freely vocalizing bats. Of greatest interest is a restricted tegmental region rostral and medial to the nuclei of the lateral lemniscus, corresponding to the paralemniscal zone (PLZ) of other mammals. Not only are most PLZ units responsive to acoustic stimuli, but a majority also show firing patterns correl- ated with vocalizations (but not respira- tion). Two types of units fire in bursts before each vocalization, with the firing rates inversely correlated either with the duration or the frequency of the constant- frequency component. Most interesting, however, is a third type, constituting approximately half of the audio-vocal units in the PLZ. These exhibit properties that make them ideal candidates for mediators of Doppler-shift compensation. Their spontaneous activity is strongly inhibited during each vocaliza- tion (hence the name VOC-inhibitory units). But when a copy of the emitted sound, reduced in intensity, delayed in time, and shifted in frequency to simulate an echo, is replayed to the bat overlapping its vocalization, all of these units show phasic-tonic responses to the 'echo'. This is remarkable behaviour: the units fail to respond to the bat's outgoing sound and their background activity is inhibited during the vocalization, yet they respond well to a fainter, overlapping, Doppler- shifted echo. The requirements of 'echo' delay and frequency differ for different units, with each showing a maximum response to a given delay and shift in frequency. Significantly, the ranges of delay and frequency shift are precisely those that are effective in eliciting beha- vioural Doppler-shift compensation. The pattern of frequency sensitivity of these VOC-inhibitory neurons is particularly important. They res- pond with exquisite sensitivity to changes in frequency of 100Hz or less, going from 10 to 100 per cent of maximal response with a change of only about 1.5 kHz, and then maintain maximal firing rates with a further increase of 2-3 kHz. This is quite unlike most other units studied, either auditory or audio- vocal, in that the frequency- response curves fall off steeply on the low-frequency side and rela- tively slowly on the high-frequency side. This is critical to their postu- lated role in mediating Doppler- shift compensation. Horseradish peroxidase injec- tion studies show several indirect connections between the PLZ and the nucleus ambiguus, and a high proportion of PLZ neurons ex- press glutamic acid decarbox- ylase, suggesting that their neuro- transmitter is GABA. Given these con- nections and the properties of VOC- inhibitory units, it is possible to build a model that can explain Doppler-shift compensation. An echo with a slight upward Doppler-shift excites some of the VOC-inhibitory neurons, which send inhibitory input to the nuclear ambiguus, reducing the firing rate of superior laryn- geal motorneurons. With larger Doppler- 1. Metzner, W. Nature 341, 529-532 (1989). 2. Schnitzler, H.U. and Henson, OW. in Animal Sonar Sys- tems (eds Busnel, R.G. and Fish, J.F.) 109-181 (NATO ASI Series, Plenum, New York, 1980). 3. Neuweiler, G. Z. vergl. Physio/. 67, 273-306 (1970). 4. Pollak, G.D. & Schuller, G. J. Neurophysiol. 45, 208 (1981). 5. Schnitzler, H.-U. Z vergl. Physio/. 57, 376-408 (1968). 6. Schuller, G. &Rubsamen, R.J. comp. Physio/.143, 317- 321 (1981). 7. Movchan, E.V. Neurobio/ogia 16, 737-745 (1984). 8. Gaioni, S.J., Suga, N. & Riquimaroux, H. Abs. Soc. Neuro· sci. 18, 442.7 (1988). 9. Suga, N. & Schlegel, P. Science 177, 82-84 (1972). 10. Mueller-Preuss, P. in Physiological Control of Mammalian Vocalization (ed. Newman, J.D.) 245-261 (Plenum, New York, 1989). 11. Schuller, G. J. comp. Physiol. 132, 39-46 (1979). NATURE · VOL 341 · 12 OCTOBER 1989